When Everybody Has the Data
In baseball, once every team had the same data, the numbers stopped being the edge. The analytics department could produce answers all day. The general manager still made the calls.
Now every business leader is getting an analytics department of their own.
Every function gets one
Marketing, sales, operations, HR, finance. All of them suddenly have access to analysis that used to take a team and a week, and now takes a good prompt and a few minutes.
Forecasts, pipeline analysis, campaign performance, turnover patterns, client profitability, the scenario where your biggest client walks. And every company is getting it, including your competitors, including the three-person shop down the street that couldn’t afford an analyst last year.
Everybody has an analytics department now. The only question left is whether the people running the business know how to use it.
Same data, two different Mondays
Say you run a 40-person agency. You hand your year-to-date numbers to AI and ask what’s going on.
Four minutes later you get back a clean, correct analysis. Gross margin is down three points since spring. Most of the drop is concentrated in two accounts, both running well past scope, and one of them is your biggest client. Your senior team is working over 100 percent of capacity. Receivables on that big account have stretched from about 45 days to about 70.
Every word of that is right. None of it tells you what to do on Monday.
That’s the analytics department’s job, done well. The leader’s job hasn’t started.
A founder who has been in that room before reads the exact same output and sees something different. The big client isn’t a margin problem. It’s a scope problem, and the slow payments are the tell, because a client who knows they’re getting more than they pay for is rarely in a hurry to pay. Don’t touch headcount. Reopen the scope at renewal, and in the meantime move one senior person off the account so your most expensive people stop quietly subsidizing it. Have the conversation before the next invoice goes out, not after the one after that.
Same data. One version is a report. The other is a decision.
AI got you the first part in four minutes, and that’s a genuine gift. The second part took somebody who has sat across from that kind of client before and knows how the conversation goes.
Questions with answers, and questions that need you
The rule for dividing the work is simple. Some questions have answers. Some need you.
Questions with answers. Hand these over.
Which clients made money last quarter, and which ones quietly cost you.
Which campaign actually drove the leads, and which one mostly drove the dashboard.
Your win rate by segment, and the stage where deals actually go to die.
Which teams are losing people, and how fast.
How many months of cash you have at the current burn.
What happens to next year if your biggest client leaves.
Every one of those has a right answer sitting in your data. AI will find it faster than you, and it won’t get bored on the fourth tab.
Questions that need you. Keep these.
Whether the client who pays on time and never complains is getting ready to leave.
Whether your best account lead is about to take a recruiter’s call.
Which partner is ready for more, and which one only thinks he is.
And whether the prospect across the table is going to be a nightmare.
AI will happily forecast the revenue from a new client. It has absolutely no idea the client is completely unhinged. You figured that out five minutes into the first call, somewhere around the third time he said he “doesn’t normally do this, but.”
That knowledge doesn’t live in any system. It lives in you, and it’s worth more than the forecast.
How to tell which is which
The line isn’t always obvious. Three quick tests.
Would two smart people with the same numbers land in the same place? If yes, the question has an answer, so hand it over. If two sharp people could look at identical data and reasonably disagree, it needs you.
Does the answer depend on something nobody ever wrote down? How the client sounded on the last call. What your partner said over dinner. The fact that your best account lead stopped cracking jokes in team meetings a month ago. If the deciding fact isn’t in any system, the machine can’t see it.
How would you find out you were wrong? When AI gets a question with an answer wrong, you can check its work against the source in ten minutes. When a judgment call goes wrong, you find out six months later, usually in somebody’s resignation letter. The slower the feedback, the more it needs a person who has seen that movie before.
Two ways to blow it
Leaders are going to get this wrong in one of two ways.
The first group will treat it as a novelty. They’ll try it once, get an answer that’s ninety percent right, decide it can’t be trusted, and go back to rebuilding the spreadsheet by hand every Thursday.
The second group will hand it the questions that need a human. Should I promote her? Is this a good client for us? Is my leadership team aligned? They’ll get back a confident, beautifully formatted answer to a question the machine has no business answering, because it has never once been in the room.
AI will answer anything you ask it with complete confidence. That’s a wonderful quality in an instrument and a terrifying one in a new hire.
The leaders who get it right will understand the output well enough to use it, and be secure enough to overrule it.
The call is still yours
Every company is about to have its own analytics department, fast, cheap and excellent, and available to every leader in the building. That part is settled. What won’t be evenly distributed is the judgment to know which questions to hand over and which ones to keep.
Use the machine for every question that has an answer, and spend the time it gives back on the ones that don’t.
The call is still yours.
Want Monday Morning Business Tips in your inbox first? Sign up at the bottom of finstratcfo.com.
Thanks for reading FinStrat CFO! Subscribe for free to receive new posts and support my work.
Originally published on Substack.
Comments